JP2001329877A - Control device for diesel engine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve precision of a learning value of an error in intake air flow under learning permitting condition and to provide a learning correction factor corresponding to difference in sensitivity to EGR under learning value reflection permitting condition. SOLUTION: A target intake air flow is calculated by a calculation means 64 based on a detected value for operating condition. An actual intake air flow is calculated by a calculation means 65. In this case, one learning value is calculated by a calculation means 66 based on the difference in the error of the intake air flow between the actual intake air flow and the target intake air flow under the learning permitting condition, while a learning value of a parameter having an influence on the sensitivity to EGR is calculated by a calculation means 67. The learning correction factor under the learning value reflection permitting condition is calculated by a calculation means 68 based on a result of comparison between the learning value of the parameter having the influence on the sensitivity to EGR and a present value thereof and also based on the leaning value of the error in the intake air flow, and an EGR device 62 is controlled by a control means 69 using a target value for controlling the EGR device 62 and the learning correction factor under the learning value reflection permitting condition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control device for a diesel engine, and more particularly to a control device including an EGR device (device for recirculating a part of exhaust gas to an intake passage) and a supercharger.

[0002]

2. Description of the Related Art When a catalyst such as a NOx reduction catalyst is provided in an exhaust passage, if the catalyst deteriorates, the exhaust pressure increases, and the exhaust pressure increases and the operation variation of a variable-capacity turbocharger increases. When the EGR amount increases due to the fluctuation of the intake air amount, the excess air ratio decreases and the smoke characteristics deteriorate. Therefore, feedback control is performed so that the actual intake air amount matches the target intake air amount (Japanese Patent Laid-Open No.
In some cases, feedback control is performed so that the actual EGR amount matches the target EGR amount (see JP-A-11-50917).

[0003]

However, the mere feedback control alone does not provide sufficient target follow-up during transients, and also complicates the setting of the feedback gain.

In addition, the relationship between the deviation of the actual intake air amount from the target intake air amount and the EGR valve opening degree is not linear, but varies depending on the differential pressure before and after the EGR valve and the EGR flow rate at that time. Therefore, it is necessary to change the feedback gain, and depending on the condition, a large amount of feedback is generated, and it is difficult to secure sufficient control accuracy.

[0005] For this reason, various experiments were conducted. As a result, the cylinder EGR amount Qec, the target EGR rate Megr, and the flow rate of the EGR gas (gas flowing through the EGR valve) (the EG flow rate).
The flow rate of R gas is hereinafter simply referred to as “EGR flow rate”.
e, a new correlation is found that is independent of the nozzle opening of the variable nozzle, and the EGR flow rate (or the EGR having a fixed relation to this flow rate) is determined based on the target EGR amount and the target EGR rate. The EGR valve is controlled based on the predicted value, and a learning value of a difference (error amount) and a ratio (error ratio) between the actual intake air amount and the target intake air amount is calculated. By correcting the EGR flow rate or the parameters (the target EGR amount and the target EGR rate) used for calculating the EGR flow rate based on the learning value of the air amount error, the exhaust pressure rise due to the deterioration of the catalyst and the variable capacity turbocharging are performed. Proposed a method that reliably prevents the excess air ratio from lowering even if the intake air amount fluctuates due to operation variations when using a machine, and further improves the robustness to exhaust gas (Japanese Patent Application See JP 11-233152).

Thereafter, it was found that the state of the EGR amount affects the controllability of the actual intake air amount to the target intake air amount.
This is because when the EGR amount is large, the actual intake air amount (or the actual intake air pressure) does not change so much when it is changed, but when the EGR amount is small, the actual intake air amount changes greatly. This is due to the point that can be made. The degree of the change in the intake air amount with respect to the EGR is called sensitivity, and when the EGR amount is large as described above, E
The usage is such that the sensitivity to GR is low, and the sensitivity to EGR increases when the EGR amount becomes small.

In this case, if the learning value of the air amount error is calculated in a region where the sensitivity to EGR is small, the learning error increases. Specifically, in the above-mentioned prior application, the rotational speed Ne and the target fuel injection amount Qsol
Is divided into a plurality of regions, and the learning value of the air amount error is stored independently for each of the divided regions. For this reason, it is basically necessary to calculate the learning values of all the regions. Therefore, whenever the learning permission condition is satisfied while the learning value reflection permission condition is satisfied, feedback correction is performed for each region where the learning permission condition is satisfied. The learning value of the air amount error is calculated based on the coefficient. Specifically, the feedback correction coefficient Kqac0,
The error ratio learning value Rqac is calculated.

A. Feedback permission condition: The error ratio Rqac0 of the air amount from the target is calculated from Equation 21 described below based on the actual intake air amount Qac and the target intake air amount delay processing value tQacd, and 1 is added to the error ratio Rqac0. The value is calculated as the feedback correction coefficient Kqac0 (because the control center of the correction coefficient is set to 1.0).

B. Learning permission condition: An error ratio Rqacn is obtained by subtracting 1 from the feedback correction coefficient Kqac0. Based on the error ratio Rqacn, the following equation (22) is used.
The error rate learning value Rqac is calculated by the equation.

When the actual intake air amount Qac is smaller than the target tQacd, for example, the feedback correction coefficient Kqac0 becomes larger than 1, and the learning value Rqac becomes a positive value of 1.0 or less. In this case,
In the region where the sensitivity to EGR is high, the actual intake air amount Q
To increase ac, the feedback correction coefficient Kqac
When 0 is greater than 1, the actual intake air amount Qac
Increases and matches the target tQacd,
In the region where the sensitivity to EGR is small, the actual intake air amount Q
To increase ac, the feedback correction coefficient Kqac
0, the feedback correction coefficient Kqac0 is further increased because the actual intake air amount Qac does not increase easily, and the learning value R calculated based on this is increased.
qac increases. That is, the learning value Rqac is calculated to be larger in the region where the sensitivity to EGR is smaller, and the learning error is larger.

Therefore, the present invention sets the learning value Rqac of the air amount error to one, and after the learning permission condition is satisfied and the learning value Rqac is calculated once and the learning is completed, the same learning value Rqac is used. Using the value Rqac, a learning correction coefficient Kqac is calculated under the learning value reflection permission condition, and the EGR flow rate or a parameter used for calculating the EGR flow rate is corrected with the correction coefficient Kqac.

In this case, when calculating the learning correction coefficient Kqac, in addition to the learning value Rqac of the air amount error,
A comparison result between a learning value of a parameter (EGR rate, engine rotation speed) affecting the sensitivity to EGR newly introduced in the present invention and its current value is considered.

Accordingly, it is an object to improve the accuracy of the learning value Rqac of the air amount error under the learning permission condition and to provide a learning correction coefficient corresponding to the difference in sensitivity to EGR under the learning value reflection permission condition.

[0014]

As shown in FIG. 70, the first invention comprises an EGR device 61, means 62 for detecting the operating conditions (for example, rotation speed and load) of the engine, and Means 63 for calculating a control target value of the EGR device 61 according to the detected value; and calculating a target intake air amount tQac or a target intake air amount equivalent value (for example, a target intake pressure) based on the detected value of the operating condition. Means 64;
Means 65 for calculating an actual intake air amount Qac or an actual intake air amount equivalent value (for example, an actual intake air pressure); and an actual intake air amount Qac or an actual intake air amount equivalent value and the target intake air amount tQac or A means 66 for calculating one learning value based on an air amount error between the target intake air amount equivalent value, a means 67 for calculating a learning value of a parameter affecting sensitivity to EGR under a learning permission condition, and
Means 68 for calculating a learning correction coefficient Kqac in a learning value reflection permission condition based on a comparison result (difference or ratio) between a learning value of a parameter affecting sensitivity to R and its current value and a learning value of the air amount error. And means 69 for controlling the EGR device 62 using the control target value of the EGR device 62 and the learning correction coefficient Kqac under the learning value reflection permission condition.

According to a second aspect of the present invention, in the first aspect, the parameters affecting the sensitivity to the EGR are an EGR rate,
It is at least one of the fuel injection amount and the engine load.

According to a third aspect, in the first aspect, the parameter affecting the sensitivity to the EGR is an engine speed.

According to a fourth aspect of the present invention, when the parameter affecting the sensitivity to the EGR in the second aspect is the EGR rate, the learning value of the EGR rate is the value of E under the learning permission condition.
This is the average value of the GR rate.

According to a fifth aspect of the present invention, when the parameter affecting the sensitivity to EGR in the third aspect is the engine speed, the learned value of the engine speed is the average value of the engine speed under the learning permission condition. is there.

According to a sixth aspect, in the first aspect, the learning permission condition is satisfied and a predetermined period (for example, a predetermined time) has elapsed.
The elapsed time is defined as the learning value reflection permission condition.

According to a seventh aspect, in the first aspect, the learning permission condition for the two types of learning values, that is, the learning value of the air amount error and the learning value of a parameter that affects the sensitivity to the EGR, is the same.

According to an eighth aspect of the present invention, in the seventh aspect, the comparison result between the learned value of the parameter affecting the sensitivity to EGR and its current value is the learning value of the parameter affecting the sensitivity to EGR and its current value. Is the difference between

In a ninth aspect of the present invention, when the parameter affecting the sensitivity to the EGR is the EGR rate in the eighth aspect, the learning value of the EGR rate is changed to the E value under the learning permission condition.
The average value of the GR rate and the learning correction coefficient Kq
When ac is constituted by the sum of the learning reflection value Rqacls and 1, the learning reflection value Rqacls is determined by the EG when the current value of the EGR rate is smaller than the learning value of the EGR rate.
It is set to be larger in absolute value than when the current value of the R rate is larger than the learned value of the EGR rate.

In a tenth aspect, in the first aspect, the learning correction coefficient Kqac is corrected with a water temperature or an atmospheric pressure.

[0024]

According to the first, second, third, fourth, and sixth aspects of the present invention, the learning permission condition is set in a region where the sensitivity to EGR is high, so that the learning error is prevented from becoming large. The accuracy of the learning value of the quantity error can be improved. Also,
When the operating region is divided into a plurality of regions and the learning value of the air amount error is configured so as to have a value independently for each region, the drivability and EGR are set for each region when setting the initial value of the learning value.
Consideration must be given to stability during control and smoke characteristics, and considerable care is required. However, according to the first, second, third, fourth, and sixth aspects of the invention, the air amount error can be reduced. Since there is only one learning value, concern for initial value setting is reduced, and adaptation becomes easier.

Further, since the learning correction coefficient in the learning value reflection permission condition is calculated based on the comparison result between the learning value of the parameter affecting the sensitivity to EGR and its current value and the learning value of the air amount error, the learning value is calculated. EGR in the reflection permission condition
A learning correction coefficient corresponding to the difference in sensitivity to.

According to the fourth and fifth aspects of the present invention, an area having a certain spread is used as a learning area to obtain an average value of the EGR rate and an average value of the engine rotation speed. The frequency of learning can be increased as compared with the case of, and even if the EGR rate or the engine rotational speed changes with a small change in the operating conditions, or even if noise occurs, the influence is not affected. Further, since the steady state and the transient do not matter when calculating the learning values of the EGR rate and the rotation speed, the learning prohibition condition can be simplified.

If the learning permission condition is made different for the two learning values, the logic of the learning value reflection permission condition becomes complicated. According to the seventh aspect of the invention, the learning permission condition is different for the two learning values. And the learning value reflection permission condition are the same, so that the logic under the learning value reflection permission condition can be simplified as compared with the case where the learning permission condition is different for each learning value.

According to the eighth aspect, E is set under the learning permission condition.
Since the learning correction coefficient is determined based on the difference between the learned value of the parameter that affects the sensitivity to GR and the current value of the parameter that affects the sensitivity to EGR under conditions that reflect the learned value, the logic can be simplified.

When the current value of the EGR rate is smaller than the learned value of the EGR rate, the sensitivity to EGR is large, and when the current value of the EGR rate is larger than the learned value of the EGR rate, the sensitivity to EGR is small. According to this, a learning reflection value corresponding to the sensitivity to EGR can be given.

Since the water temperature and the atmospheric pressure also affect the sensitivity to EGR, according to the tenth aspect, a learning correction coefficient corresponding to these can be given.

[0031]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a structure for performing a so-called low-temperature premixed combustion in which a heat generation pattern is a single-stage combustion. This configuration itself is disclosed in Japanese Patent Application Laid-Open No. 8-86.
251 and the like.

The generation of NOx greatly depends on the combustion temperature, and it is effective to lower the combustion temperature to reduce it. In the low-temperature premixed combustion, in order to realize low-temperature combustion by reducing the oxygen concentration by EGR, in response to the control pressure from the pressure control valve 5, the EGR passage 4 connecting the exhaust passage 2 and the collector 3a of the intake passage 3 is operated. A diaphragm type EGR valve 6 is provided.

The pressure control valve 5 includes a control unit 4
Driven by a duty control signal from 1
As a result, a predetermined EGR rate corresponding to the operating conditions is obtained. For example, the EGR rate is set to a maximum of 100% in a low-speed low-load region, and the EGR rate is reduced as the rotational speed and the load increase. Since the exhaust gas temperature increases on the high load side, if a large amount of EGR gas is recirculated, the effect of reducing NOx decreases due to the increase in the intake air temperature,
The EGR rate is reduced stepwise because the ignition delay period of the injected fuel becomes short and premixed combustion cannot be realized.

An EGR gas cooling device 7 is provided in the EGR passage 4. This is because the water jacket 8 is formed around the EGR passage 4 and circulates a part of the engine cooling water, and the flow control valve 9 provided at the cooling water inlet 7 a and capable of adjusting the circulation amount of the cooling water. In other words, according to a command from the control unit 41, the cooling degree of the EGR gas increases as the circulation amount increases via the control valve 9.

A swirl control valve (not shown) having a predetermined notch is provided in the intake passage near the intake port to promote combustion. When the swirl control valve is closed in the low rotation and low load range by the control unit 41, the flow velocity of the intake air taken into the combustion chamber increases, and swirl is generated in the combustion chamber.

The combustion chamber is a large diameter toroidal combustion chamber (not shown). In this, the piston cavity is formed in a cylindrical shape from the crown surface to the bottom of the piston without narrowing the inlet.At the center of the bottom, a resistance is given to the swirl flowing from the outside of the piston cavity while rotating from the outside of the piston cavity in the latter half of the compression stroke. To avoid mixing, a conical portion is formed to further improve the mixing of air and fuel. The swirl generated by the above-described swirl valve or the like due to the cylindrical piston cavity that does not restrict the inlet is diffused from the inside of the piston cavity to the outside as the piston descends in the combustion process, and even outside the cavity. Swirl is maintained.

The engine is provided with a common rail type fuel injection device 10. The configuration of the common rail type fuel injection device 10 is also known (refer to the 13th Internal Combustion Engine Symposium Lecture Papers, pp. 73-77), and is outlined with reference to FIG.

The fuel injection device 10 mainly includes a fuel tank 11, a fuel supply passage 12, a supply pump 14, a common rail (accumulator) 16, and a nozzle 17 provided for each cylinder.
The fuel pressurized by the supply pump 14 is temporarily stored in the accumulator 16 via the fuel supply passage 15, and then the high-pressure fuel in the accumulator 16 is supplied to the nozzles 1 for the number of cylinders.
7 is distributed.

The nozzle 17 includes a needle valve 18, a nozzle chamber 19,
A fuel supply passage 20 to the nozzle chamber 19, a retainer 21, a hydraulic piston 22, a return spring 23 for urging the needle valve 18 in a valve closing direction (downward in the figure), a fuel supply passage 24 to the hydraulic piston 22, this passage 24 And a three-way valve (electromagnetic valve) 25 interposed in the nozzle, and the passages 20 and 2 in the nozzle.
When the three-way valve 25 is turned off (ports A and B are connected and ports B and C are shut off), the pressure receiving area of the hydraulic piston 22 is turned off. Is larger than the pressure receiving area of the needle valve 18, the needle valve 18 is in the seated state, but the three-way valve 25 is ON.
In the state (the ports A and B are shut off and the ports B and C communicate), the fuel above the hydraulic piston 22 is returned to the fuel tank 11 via the return passage 28, and the fuel pressure acting on the hydraulic piston 22 decreases. . As a result, the needle valve 18 rises and fuel is injected from the injection hole at the tip of the nozzle. Three-way valve 2
When the valve 5 is returned to the OFF state again, the high-pressure fuel in the accumulator 16 is guided to the hydraulic piston 22 and the fuel injection ends. That is, if the fuel injection start timing is adjusted by the switching timing of the three-way valve 25 from OFF to ON, and the fuel injection amount is adjusted by the ON time, and the pressure in the accumulator 16 is the same,
The fuel injection amount increases as the ON time increases. 26 is a check valve, and 27 is an orifice.

The fuel injection device 10 further includes a pressure adjusting valve 31 in the passage 13 for returning the fuel discharged from the supply pump 14 in order to adjust the pressure of the accumulator. The adjusting valve 31 opens and closes the flow path of the passage 13, and adjusts the pressure of the accumulator by adjusting the amount of fuel discharged to the accumulator 16. The fuel injection rate changes depending on the fuel pressure (injection pressure) in the accumulator 16, and the higher the fuel pressure in the accumulator 16, the higher the fuel injection rate.

A control unit 41 to which signals from an accelerator opening sensor 33, a sensor 34 for detecting the engine rotation speed and a crank angle, a sensor 35 for determining a cylinder, and a water temperature sensor 36 are input is provided. The target fuel injection amount and the target pressure of the accumulator 16 are calculated in accordance with the pressure, and the pressure regulating valve 31 is adjusted so that the accumulator pressure detected by the pressure sensor 32 matches the target pressure.
The feedback control of the fuel pressure in the pressure accumulating chamber 16 is performed via the.

The ON time of the three-way valve 25 is controlled according to the calculated target fuel injection amount.
By controlling the switching timing to N, a predetermined injection start timing according to the operating conditions is obtained. For example, the fuel injection timing (injection start timing) is delayed up to the piston top dead center (TDC) so that the ignition delay period of the injected fuel becomes longer on the low-rotation low-load side with a high EGR rate. Due to this delay, the temperature in the combustion chamber at the time of ignition is set to a low temperature state, and the generation of smoke in the high EGR rate region is suppressed by increasing the premixed combustion ratio. On the other hand, the injection timing is advanced as the rotational speed and the load increase. This is the ignition delay crank angle (a value obtained by converting the ignition delay time into a crank angle) even if the ignition delay time is constant.
Increases in proportion to the increase in the engine speed, and the low E
In order to obtain a predetermined ignition timing at the GR rate, the injection timing is advanced.

Returning to FIG. 1, a variable displacement turbocharger is provided in the exhaust passage 2 downstream of the opening of the EGR passage 4. this is,
A variable nozzle 53 driven by a pressure actuator 54 is provided at a scroll inlet of the exhaust turbine 52.
In order to obtain a predetermined supercharging pressure from the low rotation speed range, the nozzle opening (tilting state) for increasing the flow velocity of the exhaust gas introduced into the exhaust turbine 52 is provided on the low rotation speed side, and the exhaust gas is discharged without resistance on the high rotation speed side. It is introduced into the turbine 52 and controlled to the nozzle opening degree (fully opened state).

The pressure actuator 54 includes a diaphragm actuator 55 for driving the variable nozzle 53 in response to the control pressure, and a pressure control valve 56 for adjusting the control pressure applied to the actuator 55. A duty control signal is generated so that the ratio becomes a target opening ratio Rvnt obtained as described later, and the duty control signal is output to the pressure control valve 56.

Now, from the viewpoint of supercharging pressure control,
EGR control also physically fulfills the role of boost pressure control. That is, by changing the EGR amount, the supercharging pressure also changes. Conversely, when the supercharging pressure is changed, the exhaust pressure changes, so that the EGR amount also changes. Therefore, the supercharging pressure and the EGR amount cannot be controlled independently. In addition, it is somewhat a control disturbance. In addition, if one is changed, in order to ensure control accuracy, the other must be re-adapted, but after the other is re-adapted, the other must be re-adapted. With the method, it is difficult to ensure control accuracy during transition.

As described above, the supercharging pressure and the EGR amount affect each other, and when the EGR amount is changed, it is difficult to appropriately adapt the nozzle opening degree, for example, it is necessary to change the nozzle opening degree. Control unit 41 reduces the target intake air amount tQac according to the operating conditions.
Is calculated from the target intake air amount tQac, the target EGR amount, and the target EGR rate Megr, the actual EGR amount Qec and the actual EGR rate Megrd, which are the variable target values of the turbocharger operation target. 53 target opening ratio R
Set vnt.

Further, the EGR flow velocity Cqe is predicted, and the opening Aev of the EGR valve 6 is controlled based on the predicted value.

Further, the EGR flow velocity feedback correction coefficient Kqac0 is calculated so that the actual intake air amount Qac coincides with the target intake air amount delay processing value tQacd. When the learning permission condition is satisfied, the error rate learning is performed based on the feedback correction coefficient Kqac0. In addition to calculating the value Rqac (learning value of the air amount error), a learning value of a parameter affecting the sensitivity to EGR is also newly calculated, and the learning value reflection permission condition after the calculation of the error rate learning value Rqac is completed. , A learning correction coefficient Kq based on these two types of learning values.
is calculated, and E is calculated based on the learning correction coefficient Kqac.
The actual EG which is a parameter used for calculating the GR flow velocity Cqe
The R amount Qec is corrected.

More specifically, two parameters that influence the sensitivity to EGR are the target EGR rate Megr and the engine speed Ne, and the learning value Megr of the target EGR rate is used.
1 and the learning value Nel of the engine speed are calculated under the same learning permission condition as the learning permission condition of the error rate learning value Rqac. Of these, the target EGR rate learning value Megrl and the target E
The learning reflection basic value Rqacls is obtained from the difference dMegr from the GR rate Megr (current value) and the error rate learning value Rqac.
The learning reflection value correction coefficient Krqacls is calculated from the difference dNe between the engine rotation speed learning value Nel and the engine rotation speed Ne (current value), and this correction coefficient Krqac
The learning correction coefficient Kqac is calculated based on a value obtained by multiplying the basic value Rqacls by ls.

The contents of the control executed by the control unit 41 will be described with reference to the following flowchart. 3 to 49 and FIGS. 51 to 5 described later.
9, FIG. 63 and FIG. 64 show the prior application device (Japanese Patent Application No. 11-2331).
50, FIG. 60 to FIG. 62, and FIG. 65 to FIG. 69 are modified or newly added portions in the present application.

First, FIG. 3 is for calculating the target fuel injection amount Qsol. The REF signal (a reference position signal of the crank angle, which is a signal every 180 degrees for a four-cylinder engine and every 120 degrees for a six-cylinder engine) ) Is executed for each input.

At steps 1 and 2, the engine speed Ne and the accelerator opening Cl are read. At step 3, a map containing the contents of FIG. Mqdrv is calculated, and in step 4, the basic fuel injection amount Mqdrv is increased by the engine coolant temperature or the like, and the corrected value is set as the target fuel injection amount Qsol.

FIG. 5 is for calculating the opening area Aev of the EGR valve 6, and is executed every time the REF signal is input. In step 1, a target EGR amount Tqek (control target value of the EGR device) is calculated. The calculation of Tqek will be described with reference to the flowchart of FIG.

In FIG. 7 (subroutine of step 1 in FIG. 5), in steps 1 and 2, the intake air amount Qacn per cylinder and the target EGR rate Megr are calculated.

Here, the calculation of Qacn will be described with reference to the flow of FIG. 8, and the calculation of Megr will be described with reference to the flow of FIG.

First, in FIG. 8, in step 1, the engine speed Ne is read, and this engine speed N is read.
e and the intake air amount Qas0 obtained from the air flow meter
And from

[0057]

## EQU1 ## Qac0 = (Qas0 / Ne) × KCON #, where KCON # is a constant, and the intake air amount Qac0 per cylinder is calculated.

The above air flow meter 39 (see FIG. 1)
Is provided in the intake passage 3 upstream of the compressor, and delays the transport delay from the air flow meter 39 to the collector unit 3a. Therefore, in step 3, the value of Qac0 L (where L is a constant) times before is collected by the collector. The intake air amount Qacn per cylinder at the position of the inlet 3a is obtained. Then, in step 4, this Qacn is

[0059]

## EQU2 ## Qac = Qac _n-1 × (1-KIN × KVO
L) + Qacn × KIN × KVOL, where KIN: value corresponding to volumetric efficiency, KVOL: VE / NC / VM, VE: displacement, NC: number of cylinders, VM: intake system volume, Qac _n−1 : previous Qac, (The first-order lag equation) is used to calculate the amount of intake air per cylinder at the intake valve position (the amount of intake air is hereinafter abbreviated as “cylinder intake air amount”) Qac.
This is for compensating the dynamics from the collector inlet 3a to the intake valve.

The detection of the intake air amount Qas0 on the right side of Expression 1 will be described with reference to the flowchart of FIG. The flow of FIG. 9 is executed every 4 msec.

In step 1, the air flow meter 39
The power voltage Us is read, and from this Us, at step 2 in FIG.
Retrieving a voltage-flow rate conversion table containing 0
The intake air amount Qas0 Calculate d. further,
In step 3, this Qas0 Perform weighted average processing on d
The weighted average processing value is set as the intake air amount Qas0.
Set.

Next, in FIG. 11, in step 1, the engine speed Ne, the target fuel injection amount Qsol, and the engine coolant temperature Tw are read. In step 2, the basic target EGR rate Megrb is calculated by searching a map having the contents shown in FIG. 12 from the engine speed Ne and the target fuel injection amount Qsol. In this case, the basic target EGR rate is increased in a region where the engine is frequently used, that is, in a low rotation point and a low load (low injection amount), and is reduced at a high output where smoke is likely to occur.

Next, at step 3, the cooling water temperature Tw is calculated from FIG.
By searching a table containing the following information, a water temperature correction coefficient Kegr of the basic target EGR rate is obtained. tw is calculated.
Then, in step 4, from the basic target EGR rate and the water temperature correction coefficient,

[0064]

## EQU3 ## Megr = Megrb × Kegr The target EGR rate Megr is calculated by the equation of tw.

In step 5, it is determined whether or not the state of the engine is a complete explosion state. However, this complete explosion judgment
This will be described later with reference to the flow of FIG.

In step 6, it is determined whether the state is complete explosion. If the state is complete explosion, the current processing is terminated as it is, and if it is determined that the state is not complete explosion, the target EGR rate Megr is set to 0 and the present processing is performed. finish.

Thus, the EGR control is performed after the complete explosion of the engine, and the EGR is not performed before the complete explosion to secure a stable startability.

FIG. 14 is for determining the complete explosion of the engine. In step 1, the engine rotation speed Ne is read, and the engine rotation speed Ne and the complete explosion determination slice level NRPMK corresponding to the complete explosion rotation speed are determined in step 2.
Will be compared. If Ne is larger, it is determined that the explosion is complete, and the process proceeds to step 3. Here, the counter Tmrk
b and a predetermined time TRMKBP, and a counter Tmr
If kb is longer than the predetermined time, the process proceeds to step 4 and ends assuming that the explosion has been completed.

On the other hand, if Ne is smaller in step 2, the process proceeds to step 6, where the counter Tmrkb
Is cleared, and the process is terminated in step 7 assuming that it is not in the complete explosion state. Further, even if it is larger than Ne in step 2, if the counter Tmrkb is smaller than the predetermined time in step 3, the counter is incremented in step 5 and it is determined that the explosion is not complete.

Thus, when the engine speed is equal to or higher than a predetermined value (for example, 400 rpm) and this state is continued for a predetermined time, it is determined that a complete explosion has occurred.

After the calculation of the cylinder intake air amount Qacn according to FIG. 8 and the target EGR rate Megr according to FIG. 11, the process returns to step 3 in FIG.

[0072]

The required EGR amount Mqec is calculated by the following equation: Mqec = Qacn × Megr

In step 4, this Mqec is
Let IN × KVOL be the weighted average coefficient

[0074]

Rqec = Mqec × KIN × KVOL + Rq
ec _n-1 × (1-KIN × KVOL), where KIN: equivalent value of volumetric efficiency, KVOL: VE / NC / VM, VE: displacement, NC: number of cylinders, VM: intake system volume, Rqec _n-1 : Intermediate processing value (weighted average value) Rqec is calculated by the following intermediate processing value: Rqec, and using this Rqec and the required EGR amount Mqec in step 5

[0075]

## EQU6 ## Tqec = Mqec × GKQEC + Rqec
_n-1 × (1−GKQEC), where GKQEC: advance correction gain, the advance correction is performed, and the target EGR amount Tqec per cylinder is calculated. Since there is a delay in the intake system relative to the required value (that is, a delay corresponding to the capacity of the EGR valve 6 → the collector unit 3a → the intake manifold → the intake valve), steps 4 and 5 perform processing for advancing the delay. .

In step 6,

[0077]

Equation 7: Tqek = Tqec × (Ne / KCON #) /
Kqac00, where Kqac00: EGR amount feedback correction coefficient, KCON #: constant, unit conversion (per cylinder per unit time) is performed to obtain the target EGR amount Tqek. In addition,
The calculation of the EGR amount feedback correction coefficient Kqac00 will be described later (see FIG. 54).

After completing the calculation of the target EGR amount Tqek in this way, the flow returns to step 2 in FIG. 5 to calculate the EGR flow speed Cqe, and to calculate the EGR flow speed Cqe and the target EGR amount Tqek.

[0079]

[Mathematical formula-see original document] The EGR valve opening area Aev is calculated by the following equation: Aev = Tqek / Cqe. EG
The calculation of the R flow velocity Cqe will be described later (see FIG. 63).

The EGR valve opening area Aev thus obtained is converted into a lift amount of the EGR valve 6 by searching a table containing the contents of FIG. 6 in a flow (not shown), and becomes the EGR valve lift amount. Thus, the duty control signal to the pressure control valve 5 is generated, and this duty control signal is output to the pressure control valve 5.

Next, FIGS. 15 and 16 show a control command duty value Dt applied to the pressure control valve 56 for driving the turbocharger.
This is for calculating yvnt, and is executed at regular intervals (for example, every 10 msec).

When FIG. 15 is the first embodiment and FIG. 16 is the second embodiment, there is a difference between the two embodiments in the parameters used to calculate the target opening ratio Rvnt of the variable nozzle 53 (the first embodiment in FIG. 15). In one embodiment, the actual EGR amount Qe
16 and the actual EGR in the second embodiment shown in FIG.
The target opening ratio Rvnt of the variable nozzle 53 is calculated based on the ratio Megrd).

FIGS. 15 and 16 show a main routine, in which a large control flow follows the illustrated steps, and a subroutine is prepared for the processing of each step. Accordingly, the subroutine will be mainly described below.

FIG. 17 (subroutine of step 1 in FIGS. 15 and 16) is for calculating the actual EGR rate.
It is executed every time an EF signal is input. In step 1, target EGR
The rate Megr (already obtained in FIG. 11) is read, and in step 2, a time constant equivalent value Kkin for the collector capacity is calculated. The calculation of Kkin will be described with reference to the flow of FIG.

In FIG. 18 (subroutine of step 2 in FIG. 17), in step 1, the engine speed Ne, the target fuel injection amount Qsol, and the previous value of the actual EGR rate, Megrd _n-1 [%], which will be described later, are read. Of these, Ne and Qsol are searched in step 2 for a map having the contents shown in FIG.
b, and in step 3,

[0086]

## EQU9 ## Kin = Kinb × 1 / (1 + Megrd _n−1)
/ 100) is calculated by the following equation. This is E
Since the volume efficiency is reduced by GR, the correction is made accordingly.

The value obtained by multiplying the thus determined Kin by KVOL (see step 4 in FIG. 8) which is a constant corresponding to the ratio between the intake system volume and the cylinder volume in step 4 is equivalent to a time constant corresponding to the collector capacity. The calculation is performed as the value Kkin.

When the calculation of Kkin is completed in this way, the process returns to step 3 in FIG.
Using the R rate Megr,

[0089]

Equation 10: Megrd = Megr × Kkin × Ne × K
E2 # + Megrd _n-1 × (1-Kkin × Ne × KE
2 #), where Kkin: Kin × KVOL #, KE2 #: constant, Megrd _n-1 : previous Megrd, delay processing and unit conversion (per cylinder → per unit time) are performed at the same time and the intake valve is operated. The EGR rate Megrd at the position is calculated. Ne × KE2 on the right side of Equation 10
# Is a value for unit conversion. Target EGR rate Megr
This Megrd responds with a first-order lag, so this Megrd is hereinafter referred to as “actual EGR rate”.

FIG. 20 (subroutine of step 2 in FIGS. 15 and 16) is for calculating the target intake air amount tQac. In step 1, the engine speed Ne, the actual EGR rate Megrd, and the target fuel injection amount Qsol are read, and in step 2, Megrd is compared with a predetermined value MEGLV #.

Here, the predetermined value MEGLV # is a value (for example, 0.5) for determining whether or not EGR is activated.
When Megrd> MEGRLV #, it is determined that the operating range of the EGR is in effect, and the process proceeds to steps 3, 4, and 5. On the other hand, when Megrd ≦ MEGRLV #, the EGR is performed.
And the process proceeds to step 6. MEG
The reason why RLV # is not 0 is to respond to the case where there is a request to treat a small amount of EGR in the same manner as when no EGR is performed.

If it is within the operating range of EGR, step 3
A target intake air amount basic value tQacb is calculated by searching a map having the contents shown in FIG. 21, for example, from the engine rotation speed Ne and the actual EGR rate Megrd. If the engine speed is constant, the actual EGR as shown in FIG.
The target intake air amount increases as the rate increases.

In step 4, a correction coefficient ktQac of the target intake air amount is calculated by searching a map having the contents shown in FIG. 22, for example, from Ne and Qsol, and this correction coefficient is multiplied by the target intake air amount basic value. The calculated value is calculated as the target intake air amount tQac. Correction coefficient ktQac
Is for responding to a request to change the target intake air amount according to the operating conditions (Ne, Qsol).

On the other hand, if it is in the EGR non-operating range, the routine proceeds to step 6, where the target intake air amount tQac is calculated by searching a map containing the contents of FIG. 23 from Ne and Qsol, for example.

FIG. 24 (subroutine of step 3 in FIG. 15) is for calculating the actual EGR amount. In step 1, the intake air amount Qacn per cylinder (already obtained in step 3 in FIG. 8) at the position of the collector inlet 3a, the target EGR rate Megr, and the time constant equivalent value Kkin for the collector capacity are read. Of these, Qacn and Me
gr from step 2

[0096]

The EGR amount Qec0 per cylinder at the position of the collector inlet 3a is calculated by the following equation: Qec0 = Qacn × Megr.
In step 3 using kin

[0097]

## EQU12 ## Qec = Qec0 × Kkin × Ne × KE #
+ Qec _n-1 × (1-Kkin × Ne × KE #), where Kkin: Kin × KVOL, KE #: constant, Qec _n-1 : previous Qec, The delay processing and the unit conversion (per cylinder / per unit time) are performed simultaneously to calculate the cylinder intake EGR amount Qec. Ne × KE # on the right side of Expression 12 is a value for unit conversion. Since this Qec responds to the target EGR amount Tqek with a first-order lag, this Qec is hereinafter referred to as “actual EGR amount”. The above-described Qac that responds to the target intake air amount tQac with a first-order delay is hereinafter referred to as “actual intake air amount”.

FIG. 25 (subroutine of step 4 in FIG. 15) and FIG. 27 (subroutine of step 3 in FIG. 16) are for calculating the target opening ratio Rvnt of the variable nozzle 53 (FIG. 25 is the first embodiment). FIG. 27 is a second embodiment).

Here, the opening ratio of the variable nozzle 53 is
This is the ratio of the current nozzle area to the nozzle area when the variable nozzle 53 is fully opened. Therefore, the variable nozzle 5
3, the opening ratio is 100% when fully opened, and the opening ratio is 0% when fully closed. The reason why the opening ratio is adopted is to provide versatility (a value irrelevant to the capacity of the turbocharger). Of course, the opening area of the variable nozzle may be adopted.

The turbocharger of the embodiment is of a type in which the supercharging pressure is the smallest when fully opened and the supercharging pressure is the highest when fully closed. Therefore, the smaller the opening ratio, the higher the supercharging pressure. Become.

First, referring to FIG. 25 of the first embodiment, at step 1, the target intake air amount tQac and the actual E
The GR amount Qec, the engine speed Ne, and the target fuel injection amount Qsol are read.

In steps 2 and 3,

[0103]

## EQU13 ## tQas0 = (tQac + Qsol × QFG)
AN #) × Ne / KCON #, Qes0 = (Qec + Qsol × QFGAN #) × Ne
/ KCON #, where QFGAN #: gain, KCON #: constant, an intake air amount equivalent value tQas0 for setting the target opening ratio (hereinafter, this intake air amount equivalent value is referred to as “set intake air”. EGR amount equivalent value Qes0 (hereinafter, referred to as “equivalent amount”) for setting the target opening ratio.
(This EGR amount equivalent value is referred to as “set EGR amount equivalent value”.)
Is calculated. In Equation 13, tQac and Qec are Q
The reason that sol × QFGAN # is added is that the load correction can be performed on the set intake air amount equivalent value and the set EGR amount equivalent value, and the sensitivity is adjusted by the gain QFGAN #. . Ne / KCON #
Is a value for converting into an intake air amount and an EGR amount per unit time.

From the set intake air amount equivalent value tQas0 and the set EGR amount equivalent value tQes0 obtained in this way, in step 4, for example, a map having the contents shown in FIG. 26 is searched to set the target opening ratio Rvnt of the variable nozzle 53. I do.

On the other hand, in FIG. 27 of the second embodiment,
In step 1, the target intake air amount tQac and the actual EGR rate Me
grd, engine speed Ne, target fuel injection amount Qso
Then, in step 2, the set intake air amount equivalent value tQas0 is calculated by the upper equation of the above equation (13), and the set intake air amount equivalent value tQas0 and the actual EGR are calculated.
In step 3, a target opening ratio Rvnt of the variable nozzle 53 is set by searching a map having the contents shown in FIG. 28, for example, from the rate Megrd.

The characteristics shown in FIGS. 26 and 28 are set with emphasis on fuel efficiency. However, since the difference from the exhaust-oriented setting example described later is only a specific numerical value, the characteristics common to both are described first, and then the difference between the two is described. Although the characteristics in FIG. 28 are different from those in FIG. 26 on the vertical axis (the inclination from the origin indicates the EGR rate in FIG. 26), they are basically the same as those in FIG.
The explanation will be made in the section 6.

As shown in FIG. 26, in the right region where the set intake air amount equivalent value tQas0 is large, the target opening ratio is reduced as the set EGR amount equivalent value Qes0 increases. This is for the following reason. When the EGR amount increases, fresh air decreases by that amount, and when the air-fuel ratio leans to the rich side, smoke is generated. Therefore, as the EGR amount increases, it is necessary to reduce the target opening ratio and increase the supercharging pressure.

On the other hand, the supercharging effect is not so much obtained in the small left region of tQas0. In this area, tQ
The smaller the as0 is, the smaller the target opening ratio is. This is for the following reason. This is because, if the target opening ratio is increased even in this region, it is difficult to raise the exhaust pressure because it is difficult to rise.
As described above, the characteristic shown in FIG. 26 is basically determined from two different requirements. Therefore, the change in the target opening ratio differs between a case where the change in the target intake air amount is small and a case where the change is large.

The tendency of the target opening ratio represented in FIG. 26 is common to fuel efficiency and exhaust emission, and the difference between the two is a specific numerical value. In the figure, the numerical value at the position of "small" is the minimum value at which the turbocharger works efficiently, and is the same as the fuel consumption setting example and the exhaust setting setting example, for example, about 20. On the other hand, the numerical value of the position “large” is different between the two,
In the case of a setting example of emphasis on exhaust, the value is about 30.

The setting of the target opening ratio is not limited to the above. In the first embodiment, the target opening ratio is set from the set intake air amount equivalent value tQas0 and the set EGR amount equivalent value Qes0. Instead, the target opening ratio is set from the target intake air amount tQac and the actual EGR amount Qec. It doesn't matter. Further, instead of this, the target intake air amount tQa
c and the target EGR amount (Qec0). Similarly, in the second embodiment, the target opening ratio is set from the set intake air amount equivalent value tQas0 and the actual EGR rate Megrd, but instead of this, the target intake air amount tQ
The value may be set from ac and the actual EGR rate Megrd. Further, instead of this, the target intake air amount tQac and the target EGR rate Megr may be set.

FIG. 29 (the subroutine of step 5 in FIG. 15 and step 4 in FIG. 16) is a flowchart showing the operation of the variable nozzle driving pressure actuator 54 (pressure control valve 56) with respect to the target opening ratio Rvnt obtained as described above. In order to compensate for the dynamics of the diaphragm actuator 55), advance processing is performed. This is because, when the actuator of the variable nozzle 53 is a pressure actuator, there is a non-negligible response delay unlike the case of a step motor.

In step 1, the target opening ratio Rvnt is read, and the target opening ratio Rvnt and the previous expected opening ratio Cav
Compare nt _n-1 in step 2. Here, the expected opening ratio Cavnt is a weighted average value of the target opening ratio Rvnt, as described later (see step 10).

If Rvnt> Cavnt _n-1 (while the variable nozzle 53 is being moved to the open side), the process proceeds to steps 3 and 4, the predetermined value GKVNTO # is advanced, the correction gain Gkvnt, and the predetermined value TCVNTO # are advanced and corrected. Is set as a time constant equivalent value Tcvnt of
When t <Cavnt _n−1 (when moving the variable nozzle 53 to the closing side), the process proceeds to Steps 6 and 7, and
A predetermined value GKVNTC # is advanced and a correction gain Gkvnt is advanced, and a predetermined value TCVNTC # is advanced and a time constant equivalent value Tcvn of correction is obtained.
Set as t. If Rvnt and Cavnt _n-1 are the same, the process proceeds to steps 8 and 9 to maintain the preceding advance correction gain and the value corresponding to the time constant of advance correction.

The advance correction gain Gk is obtained when the variable nozzle 53 is moved to the open side and when the variable nozzle 53 is moved to the closed side.
vnt and the value Tcvnt corresponding to the time constant for advance correction, GKVNTO # <GKVNTC #, TCVNTO #
<TCVNTC #. This is the variable nozzle 53
When moving to the closing side, it is necessary to withstand the exhaust pressure. Therefore, the gain Gkvnt is increased and the time constant is reduced accordingly (the time constant equivalent value Tcvnt, which is inversely related to the time constant, is increased). It is necessary.

In step 10, the value Tcvnt corresponding to the time constant for advance correction and the target opening ratio Rvn obtained in this way are obtained.
Using t,

[0116]

Cavnt = Rvnt × Tcvnt + Cav
nt _n-1 × (1−Tcvnt), where Cavnt _n-1 : the previous opening ratio Cavnt, the expected opening ratio Cavnt is calculated, and in step 11 based on this value and the target opening ratio Rvnt,

[0117]

[Equation 15] Avnt f = Gkvnt × Rvnt− (G
kvnt-1) × Cavnt _n−1 , where Cavnt _n−1 : the previous Cavnt, advance correction is performed, and the feedforward amount Avnt of the target opening ratio is calculated. Calculate f. Step 10, 1
The advance processing itself of 1 is basically the same as the advance processing shown in steps 4 and 5 in FIG.

FIG. 30 (each subroutine of step 6 in FIG. 15 and step 5 in FIG. 16) is a feedback amount Avnt of the target opening ratio. This is for calculating fb.
In step 1, the target intake air amount tQac and the target EGR rate M
egr, engine speed Ne, target fuel injection amount Qso
1, the actual intake air amount Qac is read, and in step 2, the target EGR rate Megr is compared with a predetermined value MEGLVV #.

When Megr ≧ MEGRLV # is satisfied (E
(When it is in the GR operating range)

[0120]

The error ratio dQac from the target intake air amount is calculated by the equation dQac = tQac / Qac-1. The value of dQac is centered on 0, and becomes a positive value when Qac as an actual value is smaller than tQac as a target value, and becomes a negative value when Qac is larger than tQac.

On the other hand, when Megr <MEGRLV # (when EGR is in the non-operating range), the routine proceeds to step 3, where the error ratio dQac = 0 (that is, feedback is prohibited).

In step 5, a correction coefficient Kh of the feedback gain is calculated by searching a predetermined map from Ne and Qsol, and this value is used in step 6 for each constant (proportional constant KPB #, integral constant KIB #, differential constant K
DB #) to obtain a feedback gain K
p, Ki, and Kd are calculated, and the feedback amount Avnt of the target opening ratio is calculated using these values. fb is calculated in step 7. The method of calculating the feedback amount is a well-known PID process.

The correction coefficient Kh is determined by the operating conditions (Ne, Ne,
Qsol), which is introduced in response to a change in the appropriate feedback gain, and increases as the load and the rotation speed increase.

FIG. 31 (each subroutine of step 7 in FIG. 15 and step 6 in FIG. 16) is for performing linearization processing on the target aperture ratio. In step 1, the feedforward amount Avnt of the target opening ratio f and feedback amount Avnt fb is read, and a value obtained by adding the two in step 2 is calculated as the command opening ratio Avnt. In step 3, this command opening ratio Avnt
For example, the command opening ratio linearization processing value Ratdty is set by searching a table (linearization table) having the contents shown in FIG.

This linear processing is required when the command signal to the actuator for driving the turbocharger has a non-linear characteristic with respect to the opening ratio (or opening area) as shown in FIG. It is. For example, as shown in FIG. 33, even when the change width of the air amount (supercharging pressure) is the same, the change width of the opening area is significantly different from dA0 and dA1 in the region having a small air amount and the region having a large air amount ( However, EG
Without R). In addition, the presence or absence of EGR (“w / o” in the figure)
The change width of the opening area also changes depending on whether “EGR” indicates no EGR and “w / EGR” indicates that EGR is present. Therefore, a target intake air amount (supercharging pressure) cannot be obtained if the same feedback gain is used regardless of operating conditions.
Therefore, in order to facilitate the adaptation of the feedback gain, the correction coefficient Kh of the feedback gain according to the operating condition is introduced as described above.

FIG. 34 (each subroutine of step 8 in FIG. 15 and step 7 in FIG. 16) sets a control command value Dtyvnt which is an ON duty value (hereinafter simply referred to as “duty value”) given to the pressure control valve 56. It is for. First, in step 1, the engine speed Ne,
Target fuel injection amount Qsol, command opening ratio linearization processing value R
Atdty, a value Tcvnt corresponding to a time constant for advance correction, and a water temperature Tw are read.

In step 2, a duty selection signal flag is set. This flag setting will be described with reference to the flow of FIG. In FIG. 35, in step 1, the command opening ratio Avnt and the time constant equivalent value Tcvnt of the advance correction are read.

[0128]

## EQU17 ## Adlyvnt = Avnt × Tcvnt + A
dlyvnt _n-1 × (1−Tcvnt), where Adlyvnt _{n-1 is a} delay process according to the formula of the previous Adlyvnt, and an expected opening ratio Adlyvn
In step 3, t is calculated, and this value is compared with Adlyvnt _nM , which is the value of M (where M is a constant) times the previous expected opening ratio.

When Adlyvnt ≧ Adlyvnt _nM (when increasing or in a steady state), the operation direction command flag fvnt is set to 1 in step 4 to indicate that it is increasing or in a steady state, and otherwise, step 5 is performed. To set the operation direction command flag fvnt = 0. In step 6, Adlyvnt and Adlyvnt _nM are compared in order to separate the case of a further increase from the steady state. If Adlyvnt = Adlyvnt _nM , the duty holding flag fvnt2 =
In other cases, the duty holding flag fvnt2 is set to 0 in step S8.

Thus, the two flags fvnt, fvnt
After completing the setting of vnt2, the process returns to step 3 in FIG. 34, and the temperature correction amount Dty of the duty value is set. Calculate t.
This calculation will be described with reference to the flow of FIG.

In FIG. 36, the engine rotational speed Ne, the target fuel injection amount Qsol, and the water temperature Tw are read in step 1, and a map including, for example, the contents shown in FIG. 37 is searched in step 2 from Ne and Qsol. The exhaust temperature Texhb is calculated. Where Te
xhb is the exhaust gas temperature after the warm-up is completed. On the other hand, if the temperature of the exhaust gas is different from the temperature of the exhaust gas after the warm-up is completed during the warming-up, the water temperature correction coefficient of the exhaust gas temperature is searched by searching the table containing the contents of FIG. Ktexh Tw is calculated, and a value obtained by multiplying the basic exhaust temperature by the above value in step 4 is calculated as an exhaust temperature Texhi.

In step 5, the exhaust temperature Texhi is

[0133]

Texhdly = Texhi × KEXH # +
Texhdly _n-1 × (1-KEXH #), where KEXH #: constant, Texhdly _n-1 : previous value of Texhdly.
Calculate as ly. This is to perform delay processing for thermal inertia.

In step 6, the basic exhaust temperature Texhb is
The difference dTexh from this actual exhaust gas temperature Texhdly is calculated.
Then, from the difference dTexh in step 7, for example,
By searching a table containing the contents shown in FIG.
Temperature correction amount Dty of the duty value Calculate t. Step
Maps 6 and 7 are maps used for hysteresis described below.
(Duty  h p, Duty  h n, Duty
l p, Duty l n map) after warm-up is completed
The difference from that state.
(That is, dTexh) with a correction amount corresponding to
is there. Note that the temperature correction amount Dty Correction by t
Actuator for driving turbocharger with temperature characteristics depending on air temperature
This is a process required when using a tutor (Fig.
40).

Thus, the temperature correction amount Dty Upon completion of the calculation of t, the process returns to the step 4 of FIG.

Steps 4 to 9 in FIG. 34 perform a hysteresis process. This processing will be described earlier with reference to FIG. 45. This is because the command opening ratio linearization processing value Ra
When dtty is increasing, the upper characteristic (Duty)
l p is the command signal for fully opening the variable nozzle, Duty h
p is a command signal when the variable nozzle is fully closed, whereas the command opening ratio linearization processing value Ratd
When ty is decreasing, another lower characteristic (Duty) l n is the command signal for fully opening the variable nozzle, D
uty The h n is to use a linear characteristic) to the command signal of the variable nozzle is fully closed. In addition, Ratdty is 1
There is a region where the two characteristics are turned upside down, but this region is not actually used.

Returning to FIG. 34, in step 4, the flag fvn
Look at t1. When fvnt = 1 (that is, when the opening ratio tends to increase or is in a steady state), the process proceeds to steps 5 and 6, and for example, a map (Duty) including FIG. h p map) and a map (Duty) containing FIG. l p map) to obtain a duty value Duty when the variable nozzle is fully closed. h and duty value when the variable nozzle is fully open Set l respectively. On the other hand, when fvnt = 0 (that is, when the opening ratio tends to decrease), the process proceeds to steps 7 and 8, and for example, a map (Duty) having the contents shown in FIG. h n map) and a map (Duty) containing FIG. l n map) to obtain the duty value Duty when the variable nozzle is fully closed. h and duty value when the variable nozzle is fully open Set l respectively.

The duty value Duty at the time of fully closing the variable nozzle thus set. h, Duty value when variable nozzle is fully open In step 9 using l and the above-described command opening ratio linearization processing value Ratdty,

[0139]

[Expression 18] Dty h = (Duty h-Duty
l) × Rattdy + Duty l + Dty The command duty value is calculated based on the linear interpolation calculation using the formula t.
Value Dty Calculate h. In other words, used for linear interpolation calculation
The characteristic of the straight line is changed, the command opening ratio linearization processing value is increasing
At or in steady state and command opening ratio linear
(When the hysteresis value is decreasing)
Cis processing), the command opening ratio linearization processing value
Even if they are the same, the command opening ratio linearization processing value tends to increase
(Or steady state)
Command duty value basic value Dty h is large
Become.

In step 10, another flag fvn
Look at t2. When fvnt2 = 1 (that is, there is no change in the command opening ratio linearization processing value), the process proceeds to step 11, where D is the previous control command duty value (to be described later).
tyvnt _n-1 is set in the normal command duty value Dtyv (duty value is held), and when fvnt2 = 0 (that is, the opening ratio is decreasing), step 12 is executed.
To the latest operation value Dty Let h be Dtyv.

At step 13, an operation confirmation control process is performed. This processing will be described with reference to the flow in FIG.
In FIG. 46, in step 1, the normal command duty value D
tyv, engine speed Ne, target fuel injection amount Qso
1. Read the water temperature Tw.

The condition for entering the operation confirmation control is determined by checking the contents of steps 2, 3, 4, and 5 one by one. When all the items are satisfied, the time until the control is further executed is determined. Enter measurement. That is, Step 2: Qsol is smaller than a predetermined value QSOLDIZ # (that is, at the time of fuel cut). Step 3: Ne is smaller than a predetermined value NEDYZ # (that is, middle rotation range). Step 4: Tw is smaller than a predetermined value TWDIZ #. Step 5: When the operation check control completion flag fdiz is 0 (the operation check control is not performed yet), the operation check control counter Ctrdiz is incremented in step 6.

In step 7, the operation check control counter is compared with predetermined values CTRDIZH # and CTRDIZL #. Here, the predetermined values CTRDIZL #, CTRDIZH
# Defines the lower limit and the upper limit of the operation check control counter, respectively. CTRDIZL # is, for example, about 2 seconds, and CTRDIZH # is, for example, about 7 seconds. Therefore, from the timing when the operation check control counter matches the lower limit CTRDIZL #, the CTR whose operation check control counter is the upper limit
While it is less than DIZH #, the routine proceeds to step 9, where an operation confirmation control command duty value is set. That is, CTR
DIZH # -CTRDIZL # is the operation confirmation control execution time.

The setting of the operation confirmation control command duty value will be described with reference to the flowchart of FIG. In FIG. 47, the operation confirmation control counter Ctrdiz and the engine speed Ne are read in step 1 and Ct is read in step 2.
From rdiz-CTRDIZL # (≧ 0), for example, FIG.
8 is searched for a control pattern Duty. Set pu. This is to move the variable nozzle 53 between the fully closed position and the fully opened position in a short cycle.

In step 3, the duty value Duty is searched by searching a table containing the contents of FIG. 49, for example, from the engine speed Ne. p ne, and this Duty p ne in step 4 the above control pattern Duty The value multiplied by pu is calculated as the control command duty value Dtyvnt. As shown in FIG. 49, the control pattern Duty Duty value to multiply pu
p ne is a value corresponding to the engine rotation speed Ne. This is based on the assumption that the duty command value for confirming the opening / closing operation of the variable nozzle 53 differs depending on the engine rotation speed. For example, the variable nozzle 53 needs to be closed against the exhaust pressure, but the exhaust pressure increases as the rotation speed increases, and accordingly the duty command value is increased. Further, on the high rotation speed side, the value is reduced so as not to be adversely affected by this control.

Referring back to FIG. 46, when the operation check control counter is less than CTRDIZL # as the lower limit, the process proceeds from step 8 to step 15, where the normal command duty value Dtyv is set as the control command duty value Dtyvnt.

When the operation check control counter becomes equal to or more than CTRDIZH # as the upper limit, step 7
Then, the process proceeds to step 10, where the previous operation check control counter Ctrdiz _n-1 and the CTR as the upper limit are set.
Compare DIZH #. Ctrdiz _n-1 <CTRDI
If it is ZH #, it is determined that the operation check control counter has just reached or exceeded CTRDIZH # as the upper limit,
In order to end the operation check control, the control command duty value Dtyvnt is set to 0 in step 11. This is because the variable nozzle 53 is fully opened once at the end of the operation check control to ensure the control accuracy in the normal control. Step 1
In 2, the operation confirmation control completed flag fdiz = 1 is set,
This processing ends. Because of the flag fdiz = 1, it is not possible to proceed to step 6 and subsequent times from the next time, so that the operation confirmation control is not performed twice after the engine is started.

If the operation check control counter has not become immediately after the count becomes higher than or equal to CTRDIZH # as the upper limit,
The process proceeds from step 10 to step 14, where the operation confirmation control counter Ctrdiz = 0 is set to prepare for the next time, and then the process of step 15 is executed.

On the other hand, Qsol is a predetermined value QSOLDIZ #
The operation confirmation control is prohibited when Ne is equal to or greater than (not during fuel cut), when Ne is equal to or greater than a predetermined value NEDIZ # (high rotation range), and when Tw is equal to or greater than a predetermined value TWDZ # (after completion of warm-up). Therefore, the process proceeds from Steps 2, 3, and 4 to Step 13, sets the flag fdiz = 0, and executes the processing of Steps 14 and 15.

As described above, when the operation of the turbocharger driving actuator is unstable, particularly at low temperatures, the operation confirmation control is performed, so that the movement of the variable nozzle becomes smooth, and the turbocharger driving The operation of the actuator can be made more reliable.

The description of FIGS. 15 and 16 has been completed.

FIG. 50 shows two feedback correction coefficients Kqa used for calculating the EGR amount and the EGR flow velocity.
c00, Kqac0 and EGR flow velocity learning correction coefficient Kqac
And is executed every time a REF signal is input.

First, at step 1, the target intake air amount tQa
c, the actual intake air amount Qac, the engine rotation speed Ne, and the target fuel injection amount Qsol are read. In step 2, the target intake air amount tQac

[0154]

[Equation 19] tQacd = tQac × KIN × KVOL ×
KQA # + tQacd _n-1 × (1-KIN × KVOL ×
KQA #), where KIN: volume efficiency equivalent value, KVOL: VE / NC / VM, VE: displacement, NC: number of cylinders, VM: intake system volume, KQA #: constant, tQacd _n-1 : last Qadd The target intake air amount delay processing value tQacd is calculated by the following equation (first-order lag equation). This is because of the delay in air supply due to the presence of the intake system volume, two feedback correction coefficients Kqac00 and Kqac0, which will be described later, and a learning value Rq.
The delay processing is performed so that ac does not become large.

At step 3, various flags related to feedback are read. These settings will be described with reference to the flowcharts of FIGS. 51, 52, 53, and 65.

FIG. 51, FIG. 52, FIG. 53, and FIG.
And is executed at regular intervals (eg, every 10 msec).

FIG. 51 shows a feedback permission flag fef.
b is set. In step 1, the engine speed Ne, the target fuel injection amount Qsol, and the actual EGR rate M
egrd and the water temperature Tw are read.

The determination of the feedback permission condition is performed by checking the contents of steps 2 to 5 and 8 one by one. Feedback is permitted when all of the items are satisfied, and feedback is performed when even one is not satisfied. Ban. That is, Step 2: Megrd exceeds a predetermined value MEGRFB # (that is, the operating range of EGR). Step 3: Tw becomes a predetermined value TWFBL # (for example, 30).
Step 4: Qsol exceeds a predetermined value QSOLFBL # (no fuel cut), Step 5: Ne exceeds a predetermined value NEFBL # (not in the engine speed range where engine stalls) Step 8: When the feedback start counter Ctrfb exceeds a predetermined value TMRFB # (for example, a value of less than 1 second), the feedback permission flag fefb = 1 is set in Step 9 to permit the feedback, and otherwise, the process proceeds to Step 10. The process proceeds to set the feedback permission flag fefb = 0 to prohibit feedback.

The feedback start counter counts up when steps 2 to 5 are satisfied (step 6), and resets the feedback start counter when steps 2 to 5 are not satisfied (step 7).

FIG. 52 shows a learning value reflection permission flag felrn.
2 is set. In step 1, the engine speed Ne, the target fuel injection amount Qsol, and the actual EGR rate M
egrd and the water temperature Tw are read.

The determination of the learning value reflection permission condition is also performed in step 2
This is performed by checking the contents of .about.5 and 8 one by one. The reflection of the learning value is permitted when all of the items are satisfied, and the reflection of the learning value is prohibited when any one of them is contrary. That is, Step 2: Megrd exceeds a predetermined value MEGRLN2 # (that is, an EGR operating range). Step 3: Tw is set to a predetermined value TWLNL2 # (for example, 2
Step 4: Qsol exceeds a predetermined value QSOLLL2 # (no fuel cut), Step 5: Ne exceeds a predetermined value NELNL2 # (not in the engine speed range where engine stalls) Step 8: When the learning value reflection counter Ctrln2 exceeds a predetermined value TMRLN2 # (for example, about 0.5 seconds), a learning value reflection permission flag feln2 = 1 is set in Step 9 to permit the reflection of the learning value. Otherwise, the process proceeds to step 10, where the learning value reflection permission flag feln2 = 0 is set to prohibit the reflection of the learning value.

The learning value reflection counter is set in steps 2 to
The count is incremented when 5 is satisfied (step 6), and reset when the steps 2 to 5 are not satisfied (step 7).

FIG. 53 is for setting the learning permission flag felrn. In step 1, the engine speed Ne, the target fuel injection amount Qsol, and the actual EGR rate Megr
d, The water temperature Tw is read.

The determination of the learning permission condition includes steps 2 to 7,
This is done by checking the contents of 10 one by one, and learning is permitted when all of the items are satisfied, and learning is prohibited when even one is wrong. That is, Step 2: Megrd exceeds a predetermined value MEGRLN # (that is, the operating range of EGR). Step 3: Tw becomes a predetermined value TWLNL # (for example, 70
Step 4: Qsol exceeds a predetermined value QSOLLL # (no fuel cut), Step 5: Ne exceeds a predetermined value NELNL # (rotational range where engine stalls) Not), Step 6: Feedback permission flag fefb = 1, Step 7: Learning value reflection permission flag feldrn2 = 1, Step 10: The learning delay counter Ctrln exceeds a predetermined value TMRLN # (for example, about 4 seconds). If yes, the learning permission flag feln = 1 is set to permit learning in step 11, otherwise, the process proceeds to step 12, and the learning permission flag feln = 0 is set to prohibit learning.

Incidentally, the learning delay counter is set in step 2
It counts up when the conditions are satisfied (step 8) and resets when the conditions are not satisfied (step 9).

FIG. 65 is for setting the learning end flag felrn3. In step 1, the learning permission flag felrn is checked. When the learning permission flag feldrn = 1, the routine proceeds to step 2, where the learning execution counter Ctr is set.
ln3 is counted up, and the learning execution counter Ctrln3 after the countup is compared with a predetermined value TMRLN3 # in step 4. Learning execution counter Ctrln3
Exceeds the predetermined value TMRLN3 #, the flow advances to steps 5 and 6 to end the learning, and the learning end flag
3 = 1, the learning permission flag feldrn = 0, otherwise, in step 7, the learning end flag feldrn3 = 0. The learning execution counter Ctrln3 is reset when the learning permission flag feldrn = 0 (steps 1 and 3).

Referring back to FIG. 50, among the four flags set in this way, the feedback permission flag fefb is checked in step 4. When fefb = 1, the feedback correction coefficient Kqac0 of the EGR amount is determined in steps 5 and 6.
0 and a feedback correction coefficient Kqac0 for the EGR flow velocity are calculated. On the other hand, when fefb = 0 (when feedback is prohibited), the process proceeds from step 4 to steps 7 and 8, where Kqac00 = 1 and Kqac0 = 1.

The calculation of the EGR amount feedback correction coefficient Kqac00 will now be described with reference to the flow chart of FIG.
The calculation of the EGR flow velocity feedback correction coefficient Kqac0 will be described with reference to the flowchart of FIG.

First, in FIG. 54 (subroutine of step 5 in FIG. 50), in step 1, the target intake air amount delay processing value tQacd, the actual intake air amount Qac, the engine speed Ne, the target fuel injection amount Qsol, and the water temperature Tw are read. .

In step 2, for example, a map having the contents shown in FIG. 55 is searched from Ne and Qsol, and so on.
The GR flow rate correction gain Gkfb is calculated, and in step 3, the water temperature correction coefficient Kgfbtw of the correction gain is calculated from Tw, for example, by searching a table having the contents shown in FIG. 56.

[0171]

Kqac00 = (tQacd / Qac-1)
× Gkfb × Kgfbtw + 1, the EGR amount feedback correction coefficient Kqac0
Calculate 0.

The first term on the right side of this equation, (tQacd / Qa
c-1) is the error ratio from the target intake air amount delay processing value. By adding 1 to this, Kqac00 becomes a value centered on 1. Equation 20 calculates the EGR amount feedback correction coefficient Kqac00 in proportion to the error ratio from the target intake air amount delay processing value.

Next, in FIG. 57 (subroutine of step 6 in FIG. 50), in step 1, the target intake air amount delay processing value tQacd, the actual intake air amount Qac, the engine speed Ne, the target fuel injection amount Qsol, and the water temperature Tw. Read.

In step 2, by searching a map containing the contents of FIG. 58 from Ne and Qsol, for example,
In step 3, the correction gain Gkfbi of the GR flow velocity is calculated, and in step 3, the water temperature correction coefficient Kgfbitw of the correction gain is calculated from Tw by searching a table containing, for example, FIG.

[0175]

Rqac0 = (tQacd / Qac-1) ×
Gkfbi × kGfbitw + Rqac0 _n−1 , where Rqac0 _n−1 : the previous error rate Rqac0, the error rate Rqac0 is updated, and the value obtained by adding 1 to the error rate Rqac0 in step 5 is the EGR flow rate feedback correction coefficient Kqac0. Is calculated as

The EGR flow velocity feedback correction coefficient K is proportional to the integrated value (integral value) of the error ratio (tQacd / Qac-1) from the target intake air amount delay processing value.
qac0 is calculated (integral control).

The reason why the correction gain is set to a value corresponding to the operating condition (Ne, Qsol) as shown in FIGS. 55 and 58 is as follows. Even if the gain is the same, hunting may or may not occur depending on the operating conditions, so that the correction gain is reduced in a region where hunting occurs. FIG.
The reason why the value is reduced at the time of low water temperature (before completion of warm-up) as shown in FIG. 59 is to stabilize the engine in a low water temperature region where engine rotation is unstable.

After the calculation of the EGR amount feedback correction coefficient Kqac00 and the EGR flow velocity feedback correction coefficient Kqac0 is completed, the flow returns to FIG. 50, and the learning value reflection permission flag feldrn is determined in steps 9, 10, and 11.
2. Learning permission flag feldrn, learning end flag feld
Look at rn3.

Here, how the respective flags are set in response to changes in the operating conditions is described. For example, when the target fuel injection amount Qsol increases slowly, FIG.
It becomes as shown in. However, the target fuel injection amount Qsol
Focusing only on the above, no conditions other than Qsol are considered. In addition,
In the figure, four predetermined values (QSOLFBL #, QSO
LLNL2 #, QSOLLLN #, QSOLLLNL3
The magnitude relation of #) is not limited to this.

Referring to FIGS. 69 and 50 together, FIG.
9, different operations are performed in (1) a section from t2 to t4, (2) a section from t4 to t5, and (3) a section after t5. Hereinafter, these three cases will be described separately according to FIG.

(1) Interval from t2 to t4: The interval from t2 to t3 is a section in which the reflection of the learning value is prohibited (the learning value reflection permission flag felrn2 = 0). In this section, the reflection of the learning value is prohibited. Step 9 to Step 12 in FIG.
Then, the EGR flow rate learning correction coefficient Kqac = 1 is set.

The section from t3 to t4 is a section where the condition for permitting the reflection of the learning value has just been set (the learning value reflection permission flag felrn2 = 1 and the learning permission flag felrn).
= 0 and the learning end flag felrn3 = 0), and this section is treated the same as the section in which the reflection of the learning value is prohibited for reasons such as waiting for the state to stabilize.
Go to Step 12 from Steps 9, 10, and 11 of E
The GR flow velocity learning correction coefficient Kqac = 1.

(2) Section from t4 to t5: The section from t4 to t5 is a section where the learning permission condition is satisfied (the learning value reflection permission flag felrn2 = 1 and the learning permission flag fe
lrn = 1), and proceeds to steps 13 to 16 from steps 9 and 10 in FIG. That is, the error ratio learning value Rqac is read in step 13, and a value obtained by adding 1 to the error ratio learning value Rqac is calculated as the EGR flow rate learning correction coefficient Kqac under the learning permission condition in step 14.

In step 15, 1 is subtracted from the EGR flow velocity feedback correction coefficient Kqac0 to obtain an error rate Rq
acn, and using this and the error rate learning value Rqac,
In step 16,

[0185]

Rqac _n = Rqacn × Tclrn + Rq
ac _n-1 × (1-Tclrn), where Tclrn: learning speed, Rqac _n : error ratio learning value after update, Rqac _n-1 : error ratio learning value before update, and weighted average processing is performed. To calculate the error ratio learning value (by overwriting the value before update with the value before update).

The control in the section (2) and the control in the section (1) described here are the same as in the prior application (Japanese Patent Application No. 11-233152). However, in the prior application, the operating region using Ne and Qsol as parameters is divided into a plurality of regions, and an independent error rate learning value Rqac is stored for each of the divided regions. The error ratio learning value Rqac of the invention is composed of only one. The control under the learning value reflection condition in the section (3) described below is a part newly added in the present invention.

(3) Section after t5: The section after t5 is the error rate learning value Rqa calculated in the above section (2).
This is a section in which c is reflected in the calculation of the EGR flow velocity Cqe (the learning value reflection permission flag feldrn2 = 1, the learning permission flag feldrn = 0, and the learning end flag feldrn3).
= 1), steps 9, 10, 11 in FIG.
Proceeding to step 17, the error rate learning value Rqac,
The EGR flow rate learning correction coefficient Kqac under the learning value reflection permission condition is calculated based on two types of learning values, that is, a learning value of the EGR rate and a learning value of the engine rotation speed (a learning value of a parameter that affects sensitivity to EGR). This learning correction coefficient K
The operation of qac will be described with reference to the flow of FIG.

In FIG. 60 (subroutine of step 17 in FIG. 50), in step 1, the target EGR rate Megr,
The engine speed Ne is read. In step 2, the learning value Megrl of the target EGR rate (described later with reference to FIG. 66)
And the learning value Megrl and the target EGR rate Me
difference dMegr from gr (current value) (= Megr−Me)
grl) is calculated in step 3. In step 4, the learning value Rqacs of the error ratio is read, and a learning reflection basic value Rqacls is calculated by searching a map having the contents shown in FIG. 61 from the Rqac and the difference dMegr.

In step 6, a learning value Nel (described later with reference to FIG. 66) of the engine speed is read, and the difference d between the learning value Nel and the engine speed Ne (current value) is read.
Ne (= Ne−Nel) is calculated, and a learning reflection value correction coefficient Krqacls is calculated by searching a table having the contents shown in FIG. 62 from the difference dNe. Then, the correction coefficient Krqacls and the learning reflection basic value Rq are used.
Using acls, Kqac = Rqacls × Krqa
cls + 1 (31) The EGR flow rate learning correction coefficient in the learning value reflection permission condition after the learning is completed is calculated by the equation (31).

The characteristic shown in FIG. 61 is the error rate learning value Rqa.
The characteristic is as if the upper and lower parts were turned upside down at the line of c = 0 (the upper half is a positive value and the lower half is a negative value), and the character does not change between the upper half and the lower half. Here, the characteristic of the upper half will be described as an example. First, in a region where Rqac is positive, the value of the learning reflection basic value Rqacls is also assumed to be a positive value.
The reason why the learning correction coefficient Kqac is set to a value exceeding 1 is as follows.
This is the same reason that the learning correction coefficient Kqac is set to a value exceeding 1 in step 14 of FIG. The error ratio learning value Rqac is for making the actual intake air amount Qac coincide with the target intake air amount delay processing value tQacd. To realize this, the error ratio learning value Rqac is reflected in a positive region. This is because it is necessary to configure the value of the basic value Rqacls as a positive value. In other words, the case where the error rate learning value Rqac is positive means that the target value tQ
In this case, if the learning correction coefficient Kqac is set to a value greater than 1, the EGR flow rate Cqe is corrected to be larger as will be described later (the correction is performed using the learning correction coefficient Kqac). Actual EGR
Quantity Qec h becomes large, and E is obtained through the map of FIG.
The GR flow rate Cqe increases), whereby the EGR valve opening area Aev decreases, the actual intake air amount increases, and approaches the target tQacd.

FIG. 5 shows that the value of the learning reflection basic value Rqacls is increased so that the learning correction coefficient Kqac increases as the error rate learning value Rqac increases.
This is for the same reason that the learning correction coefficient Kqac increases as the error ratio learning value Rqac increases in step 14 of 0.

Therefore, the main difference from the learning permission condition is that as the difference dMegr of the target EGR rate increases with a positive value, the value of the learning reflection basic value Rqacls decreases, and conversely, the difference dMegr becomes a negative value. The point is that the value of Rqacls is increased as the value becomes larger. In other words, the learning reflection basic value Rqacls is set in accordance with the sensitivity to EGR. That is, in FIG. 61, the sensitivity to EGR decreases as the position shifts to the right, and conversely, the sensitivity to EGR increases as the position shifts to the left. Therefore, the value of Rqacls increases as the sensitivity to EGR increases. . For comparison, the case of the prior application device will be described. In the prior application device, since the sensitivity to EGR is not taken into consideration, the image is an image in which Rqacls has the same value (the characteristic is horizontal) regardless of the sensitivity to EGR. In addition, the point of dMegr = 0 is a learning point.

Next, referring to FIG. 62, the rotational speed difference dNe
Sets the value of the learning reflection value correction coefficient Krqacls to a value exceeding 1.0 in the positive region for the following reason. dN
e is positive when the actual rotation speed Ne (current value) is greater than the rotation speed learning value Nel.
Since the flow velocity increases (that is, the sensitivity to EGR increases), the learning correction coefficient Kqac is corrected in accordance with the increase.

Next, the calculation of the learning values Megrl and Nel of the target EGR rate and the engine speed will be described with reference to FIG.
6 will be described. In FIG. 66, in step 1, the learning permission flag feldrn is checked. Learning permission flag ferr
When n = 1, the target EGR rate Megr and the engine speed Ne (both are current values) are read in step 2.

In step 3, the learning value Me of the target EGR rate is obtained.
grl is read out, and in step 4 using this learning value Megrl and its current value Megr, Megrl = Megr × TCLNEG + Megrl _n−1 × (1−TCLNEG) (32) where Megrl: after update Target EGR rate learning value, Megrl _n-1 : target EGR rate learning value before update, TCLNEG: learning speed (constant), weighted averaging process is performed to obtain the target EGR rate learning value in the same manner. In steps 5 and 6, the learning value Nel of the engine rotation speed is read out, and using this learning value Nel and its current value Ne, Nel = Ne × TCLNNE + Nel _n−1 × (1-TCLNNE) (33) However, Nel: rotation speed learning value after the update, Nel _n-1: rotation speed learning value of the previous update, TCLNNE: learning speed (constant), the expression of The weighted average processing each for calculating a learning value of the engine rotational speed by performing (overwriting the updated values with respect to the pre-update value).

Here, as shown in FIGS. 61 and 62, to set the learning reflection basic value Rqacls and the learning reflection value correction coefficient Krqacls, the difference dMegr, dN
Since e is required and the difference is the difference between the learned value and its current value, it is necessary to obtain (learn) a learned value. In this case, as one method, it is conceivable that the learning region is limited to a predetermined operation region and learning is performed in a steady state. However, this method is not preferable because the frequency of learning is reduced in the case of an automobile diesel engine whose operating conditions change frequently. On the other hand, in the present embodiment, the above (3)
Since the weighted average value is configured as the learning value as shown in the equations (2) and (33), the learning is performed regardless of the steady state or the transient state, and the learning frequency does not decrease. However, to prevent a large error from occurring in the learning value (32)
Learning speed TCLNEG, TCLNNE in equation (33)
It is necessary to determine.

FIG. 63 (subroutine of step 2 in FIG. 5) is for calculating the EGR flow velocity Cqe.

In steps 1 and 2, the actual EGR amount Qec and the actual EGR
The GR rate Megrd, the actual intake air amount Qac, the EGR flow velocity feedback correction coefficient Kqac0, and the EGR flow velocity learning correction coefficient Kqac are read.

[0199]

## EQU23 ## Qec From the equation of h = Qec × Kqac × Kqac0, the actual EGR amount Qec is calculated by Kqac0 and Kqac.
The corrected actual EGR amount Qec h, and the corrected actual EGR amount Qec h and actual EGR rate Meg
In step 8, the EGR flow speed Cqe is calculated by searching a map having the contents shown in FIG. 64 from rd. Steps 4 to 7, which have not been described, will be described later.

Since the characteristic of the EGR flow velocity in FIG. 64 shows that the nonlinearity is strong and the sensitivity of the EGR feedback differs depending on the operating condition, the difference in the feedback amount with respect to the operating condition is reduced. The EGR flow velocity feedback correction coefficient Kqac0 is used as feedback to the actual EGR amount Qec used for searching the flow velocity map.

However, in FIG. 64, the portion near the right end where the slope of the characteristic is steep is a region in which a matching error of the map tends to occur, and if there is a matching error, the EGR valve is affected by the matching error. The opening area Aev changes.
That is, A is a formula for calculating the EGR valve opening area Aev.
At ev = Tqek / Cqe, a matching error occurs in Cqe.
It is necessary to correct the flow rate error for the amount Tqek. Therefore, the newly introduced EGR amount feedback correction coefficient Kqac00 is used to correct the target EGR amount Tqek in step 6 of FIG. 7 using the Kqac00.

In this case, the above equation (20) for calculating Kqac00 calculates Kqac00 in proportion to the error ratio from the target intake air amount delay processing value. Therefore, the EGR flow rate map shown in FIG. Can be immediately corrected for the matching error. For example, in Equation 20, for simplicity, considering that the correction gain Gkfb = 1 and the warm-up is completed, Kqac00 = (tQacd / Q
ac-1) +1. In this case, t as the target value
If the actual intake air amount Qac is smaller than Qacd, Kqac
00 becomes a value greater than 1 and Tqec is immediately reduced. When the target EGR amount is immediately reduced, the fresh air amount (intake air amount) relatively increases, whereby the actual intake air amount Qac converges to tQacd as the target value.

Steps 4 to 7 in FIG. 63, which have not been described, are for setting initial values at the start of EGR operation. Specifically, in step 4, the corrected actual EGR amount Qec Compare h with 0. Qec When h = 0 (that is, when EGR is not operating), the process proceeds to step 5, and

[0204]

[Equation 24] Qec h = Qac × MEGRL #, where MEGRL #: constant, and the corrected actual EGR amount Qec Set h. Similarly, in step 6, the actual EGR rate Megrd is compared with 0, and when Megrd = 0, in step 7,

[0205]

## EQU25 ## The actual EGR rate Megrd is set according to the following equation.

When the EGR valve 6 is fully closed, the flow rate of the EGR passing through the EGR valve 6 is naturally zero.
Equation (25) sets the initial values of the parameters used for the calculation of the flow velocity in consideration of the start of the EGR operation. ME
The value of GRL # is, for example, 0.5 as described above.
More specifically, the differential pressure before and after the EGR valve (and hence the EGR flow rate) at the start of the EGR operation differs depending on the operating conditions, and this is dealt with. In this case, EG
The differential pressure across the EGR valve at the start of R operation is the actual intake air amount Q
related to ac. Therefore, according to equation 24, Qec is proportional to Qac. By giving the initial value of h, the calculation accuracy of the EGR flow velocity at the start of the EGR operation is improved.

Here, the operation of the two embodiments will be described.

In the two embodiments, the target of learning correction is the EGR flow rate Cqe, which is different from the prior art. In this case, the variable nozzle is located between the cylinder EGR amount Qec, the target EGR rate Megr, and the EGR flow rate Cqe. Based on the first finding that there is a strong correlation regardless of the nozzle opening, the EGR flow rate Cqe is predicted as in the present embodiment, and the EGR valve opening is controlled based on the predicted value. Even in an engine having a variable capacity turbocharger, the target EGR amount can be accurately calculated irrespective of the nozzle opening of the variable nozzle, and the EGR gas flow rate Cqe is a value irrespective of steady state and transient. Therefore, the target EGR amount including the transient can be accurately calculated.

The above is a description of the prior application (Japanese Patent Application No. 11-23315).
2), and the operation of the part added according to the present invention will now be described.

In this embodiment, one error ratio learning value Rqac (air value error learning value) is calculated under the learning permission condition (see steps 15 and 16 in FIG. 50), and EGR is performed.
Affecting target sensitivity (target EGR rate Me
gr, the engine rotation speed Ne) is calculated (FIG. 6).
6), when the learning value reflection permission condition after the end of the calculation of the learning value is satisfied, the learning value (Megrl, Nel) of the parameter affecting the sensitivity to EGR and its current value (Meg)
r, Ne) and the error ratio learning value Rqac
A learning correction coefficient Kqac is calculated on the basis of the learning correction coefficient Kqac (see steps 9, 10, 11, 17, and FIG. 60 in FIG. 50), and the actual EGR amount Qec, which is a parameter used for calculating the EGR flow velocity, is corrected using the learning correction coefficient Kqac. (See step 3 in FIG. 63).

That is, according to the present embodiment, since there is only one error rate learning value Rqac, this learning value Rqac
If the learning permission condition is set in a region where the sensitivity to EGR is large, the learning value Rq
The accuracy of ac can be improved. In addition, when the operating region is divided into a plurality of regions and the learning value Rqac is configured to have a value independently for each region, the drivability and EGR control are performed for each region when setting the initial value of the learning value Rqac. Consideration must be given to stability and smoke characteristics at the time, and considerable care is required. However, according to the present embodiment, since the learning value Rqac is one, care for initial value setting is reduced. It is easier to fit.

The learning correction coefficient Kqac under the learning value reflection permission condition is calculated based on the comparison result between the learning value of the parameter affecting the sensitivity to EGR and its current value and the error ratio learning value Rqac (FIG. 60). In the learning value reflection permission condition, a learning correction coefficient Kqac corresponding to a difference in sensitivity to EGR can be given.

Parameters that affect the sensitivity to EGR are the target EGR rate and the engine speed, and the learning values of the target EGR rate and the rotation speed are the average values of the target EGR rate and the rotation speed under the learning permission condition. . In order to obtain the average value of the target EGR rate and the average value of the engine rotation speed, the region having a certain width is the learning region. In addition, even if the target EGR rate or the engine rotation speed changes due to a slight change in the operating conditions, or even if noise occurs, the influence is not affected. In addition, since the calculation of the learning values of the target EGR rate and the rotation speed does not take into consideration the steady state and the transient state,
Learning prohibition conditions can be simplified.

Further, the error ratio learning value Rqac and the target EG
If the learning permission condition is made different for the two learning values of the learning rate Megrl and Nel of the R rate and the rotation speed, the logic under the learning value reflection permission condition becomes complicated. Since the learning permission condition and the learning value reflection permission condition are the same for the two types of learning values, compared to the case where the learning permission conditions are different for the two types of learning values,
The logic under the learning value reflection permission condition can be simplified.

The difference dMe between the target EGR rate and the learning value Megrl and Nel of the rotation speed under the learning permission condition and the current value of the target EGR rate and the rotation speed under the learning value reflection permission condition.
Since the learning correction coefficient Kqac is determined by gr and Ne, the logic can be simplified.

When the difference dMegr of the target EGR rate is negative (when the current value of the EGR rate is smaller than the learning value of the EGR rate), the sensitivity to EGR is large, and when dMegr is positive (the EGR rate is smaller). Is larger than the learning value of the EGR rate), the learning reflection basic value Rqacls is set to correspond to the low sensitivity to EGR, and the target EG is set to be smaller when the difference dMegr of the target EGR rate is negative.
Since the difference dMegr of the R rate is set to be larger in absolute value than when it is positive, the learning reflection basic value Rqacls corresponding to the sensitivity to EGR can be given.

Although the embodiment has been described in consideration of the influence on the EGR sensitivity due to the difference between the target EGR rate and the rotation speed, the water temperature Tw and the atmospheric pressure Pa also affect the EGR sensitivity via the EGR rate. , It is desirable to introduce correction by these. For example, instead of the above equation (31), the EGR flow rate learning correction coefficient in the learning value reflection permission condition is calculated by the following equation: Kqac = Rqacls × Krqacls × water temperature correction coefficient × atmospheric pressure correction coefficient + 1 (34) In step 14 of FIG. 50, the EGR flow rate learning correction coefficient in the learning permission condition is calculated by the following equation: Kqac = Rqac × water temperature correction coefficient × atmospheric pressure correction coefficient + 1 (35).

The water temperature correction coefficients of the equations (34) and (35)
The characteristics of the atmospheric pressure correction coefficient are shown in FIGS. Figure 67
The reason why the value becomes larger as the water temperature becomes lower in the above is that the target EGR rate becomes smaller (the sensitivity to EGR becomes larger) at the lower water temperature. In FIG. 68, the value is increased at high altitude because the target EGR rate decreases at high altitude (the sensitivity to EGR increases).

In the embodiment, the learning correction coefficient Kqac under the learning value reflection permission condition is calculated based on the difference between the learning value of the parameter affecting the sensitivity to EGR and its current value and the learning value of the air amount error. As explained, EG
The learning correction coefficient Kqac under the learning value reflection permission condition may be calculated based on the ratio between the learning value of the parameter affecting the sensitivity to R and its current value and the learning value of the air amount error.

In the embodiment, the case where the learning permission condition of the learning value of the air amount error and the learning permission condition of the learning value of the parameter affecting the sensitivity to EGR are the same is described. I don't care.

In the embodiment, the case has been described where the parameters affecting the sensitivity to EGR are the target EGR rate and the rotation speed. However, either one of them may be introduced. Further, the present invention is not limited to the target EGR rate. For example, since the target EGR rate is determined by the target fuel injection amount Qsol, the fuel injection amount Qsol may be used or the engine load may be used. The actual EGR rate can also be used.

In the embodiment, when the parameters affecting the sensitivity to EGR are the EGR rate and the engine speed, the learned values of the EGR rate and the engine speed are calculated based on the weighted average of the EGR rate and the engine speed under the learning permission condition. Although the description has been made in the case of the value, the value is not limited to the weighted average value.

In the embodiment, the case where the EGR flow rate is predicted and the EGR valve is controlled based on the predicted value has been described. (Proportional to the square of.).

Further, using the target opening ratio Rvnt and the time constant equivalent value Tcvnt, RVNTE = Rvnt × Tcvnt × KVN1 # + RVNTE _n−1 × (1−Tcvnt × KVN1 #) (36) where KVN1 # : Constant, RVNTE _n-1 : previous RVNTE, there is one that calculates the delay processing value RVNTE of the target opening ratio by the formula (see Japanese Patent Application No. 2000-139929). EG
It can be adopted as a value that approximates the differential pressure across the R valve. When the EGR valve opening degree is constant, RVNTE
Is smaller, the variable nozzle 53 is closed and the supercharging pressure is increased, so that the differential pressure across the EGR valve is increased. Conversely, as RVNTE is increased, the variable nozzle 53 is opened and the supercharging pressure is decreased, so that the EGR valve The differential pressure before and after becomes smaller. If the RVNTE and the EGR amount are used as parameters, the opening area of the EGR valve can be directly given. The characteristic at this time is that if the EGR amount is constant, the RV
NTE increases (the differential pressure across the EGR valve decreases)
The EGR valve opening area becomes smaller as the RVNTE becomes constant, and the opening area becomes larger as the EGR amount increases. When the EGR valve is controlled by such a configuration, the target EGR set based on the detected value of the operating condition is set.
The amount may be corrected by the learning correction coefficient under the learning value reflection permission condition, and the opening area of the EGR valve may be obtained using the corrected target EGR amount.

In the present embodiment, the EG is controlled so that the actual intake air amount Qac matches the target intake air amount delay processing value tQacd.
An R flow velocity feedback correction coefficient Kqac0 is calculated, and the EGR flow velocity C is calculated using the feedback correction coefficient Kqac0 and the EGR flow velocity learning correction coefficient Kqac (= Rqac + 1).
The case where the cylinder EGR amount Qec which is a parameter used in the calculation of qe is corrected has been described, but the actual EGR rate M which is another parameter is the EGR flow velocity feedback correction coefficient Kqac0 and the EGR flow velocity learning correction coefficient Kqac.
Even if the egrd is corrected, the actual intake air amount Q
EGR so that ac matches the target intake air amount tQac.
The flow rate feedback correction coefficient Kqac0 is calculated, and the feedback correction coefficient Kqac0 and the EGR flow rate learning correction coefficient Kqac are used to correct the cylinder EGR amount Qec and the actual EGR rate Megrd used in the calculation of the EGR flow rate Cqe, and further, the EGR flow rate Cqe itself. It doesn't matter. As for the learning value, the learning correction coefficient may be used as the learning value.

The target intake air amount and the actual intake air amount of the present invention are not limited to the target intake air amount tQac and the actual intake air amount Qac described in the embodiment, and may be the intake air amount such as the target intake pressure and the actual intake pressure. A corresponding value may be used.

In the embodiment, the case of performing so-called low-temperature premix combustion in which the pattern of heat generation is single-stage combustion has been described. However, in the case of normal diesel combustion in which diffusion combustion is added after premix combustion, However, it goes without saying that the present invention can be applied.

[Brief description of the drawings]

FIG. 1 is a control system diagram of an embodiment.

FIG. 2 is a schematic configuration diagram of a common rail type fuel injection device.

FIG. 3 is a flowchart for explaining calculation of a target fuel injection amount.

FIG. 4 is a map characteristic diagram of a basic fuel injection amount.

FIG. 5 is a flowchart for explaining the calculation of an EGR valve opening area.

FIG. 6 is a characteristic diagram of an EGR valve drive signal with respect to an EGR valve opening area.

FIG. 7 is a flowchart for explaining calculation of a target EGR amount.

FIG. 8 is a flowchart for explaining calculation of a cylinder intake air amount.

FIG. 9 is a flowchart for explaining detection of an intake air amount.

FIG. 10 is a characteristic diagram of an intake air amount with respect to an air flow meter output voltage.

FIG. 11 is a flowchart for explaining calculation of a target EGR rate.

FIG. 12 is a map characteristic diagram of a basic target EGR rate.

FIG. 13 is a table characteristic diagram of a water temperature correction coefficient.

FIG. 14 is a flowchart illustrating a complete explosion determination.

FIG. 15 is a flowchart for explaining calculation of a control command duty value given to the pressure control valve according to the first embodiment.

FIG. 16 is a flowchart for explaining calculation of a control command duty value given to the pressure control valve according to the second embodiment.

FIG. 17 is a flowchart for explaining calculation of an actual EGR rate.

FIG. 18 is a flowchart for explaining calculation of a time constant equivalent value for a collector capacitance.

FIG. 19 is a map characteristic diagram of a basic value corresponding to volume efficiency.

FIG. 20 is a flowchart for explaining calculation of a target intake air amount.

FIG. 21 is a map characteristic diagram of a target intake air amount basic value during EGR operation.

FIG. 22 is a map characteristic diagram of a target intake air amount correction coefficient.

FIG. 23 is a map characteristic diagram of a target intake air amount when EGR is not operated.

FIG. 24 is a flowchart for explaining calculation of an actual EGR amount.

FIG. 25 is a flowchart illustrating the calculation of a target opening ratio according to the first embodiment.

FIG. 26 is a map characteristic diagram of a target opening ratio.

FIG. 27 is a flowchart illustrating a calculation of a target opening ratio according to the second embodiment.

FIG. 28 is a map characteristic diagram of a target opening ratio.

FIG. 29 is a flowchart for explaining the calculation of the feedforward amount of the target opening ratio.

FIG. 30 is a flowchart for explaining calculation of a feedback amount of a target opening ratio.

FIG. 31 is a flowchart illustrating a linearization process.

FIG. 32 is a table characteristic diagram of linearization.

FIG. 33 is a characteristic diagram showing a relationship between an opening area and a supercharging pressure.

FIG. 34 is a flowchart for explaining signal conversion.

FIG. 35 is a flowchart illustrating the setting of a duty selection signal flag.

FIG. 36 is a flowchart for explaining the calculation of the temperature correction amount of the duty value.

FIG. 37 is a map characteristic diagram of a basic exhaust gas temperature.

FIG. 38 is a table characteristic diagram of a water temperature correction coefficient.

FIG. 39 is a table characteristic diagram of a temperature correction amount.

FIG. 40 is a temperature characteristic diagram of an actuator for driving a turbocharger.

FIG. 41 is a map characteristic diagram of the duty value when the variable nozzle is fully closed.

FIG. 42 is a map characteristic diagram of the duty value when the variable nozzle is fully opened.

FIG. 43 is a map characteristic diagram of the duty value when the variable nozzle is fully closed.

FIG. 44 is a map characteristic diagram of the duty value when the variable nozzle is fully opened.

FIG. 45 is a hysteresis diagram when a command opening ratio linearization processing value is converted into a duty value.

FIG. 46 is a flowchart for explaining operation confirmation control;

FIG. 47 is a flowchart for describing the setting of the operation check control command duty value.

FIG. 48 is a table characteristic diagram of a control pattern.

FIG. 49 is a table characteristic diagram of duty values during operation check control.

FIG. 50 is a flowchart for explaining the calculation of two feedback correction coefficients and a learning correction coefficient of EGR control.

FIG. 51 is a flowchart illustrating the setting of a feedback permission flag.

FIG. 52 is a flowchart for describing setting of a learning value reflection permission flag.

FIG. 53 is a flowchart illustrating the setting of a learning permission flag.

FIG. 54 is a flowchart illustrating the calculation of an EGR amount feedback correction coefficient.

FIG. 55 is a map characteristic diagram of a correction gain of the EGR flow rate.

FIG. 56 is a table characteristic diagram of a water temperature correction coefficient.

FIG. 57 is a flowchart for explaining the calculation of an EGR flow velocity feedback correction coefficient.

FIG. 58 is a map characteristic diagram of a correction gain of the EGR flow velocity.

FIG. 59 is a table characteristic diagram of a water temperature correction coefficient.

FIG. 60 is a flowchart for explaining the calculation of an EGR flow velocity learning correction coefficient under a learning value reflection permission condition.

FIG. 61 is a map characteristic diagram of a learning reflection basic value.

FIG. 62 is a table characteristic diagram of a learning reflection value correction coefficient.

FIG. 63 is a flowchart for explaining the calculation of the EGR flow velocity.

FIG. 64 is a map characteristic diagram of the EGR flow velocity.

FIG. 65 is a flowchart for describing the setting of a learning end flag.

FIG. 66 is a flowchart for explaining calculation of each learning value of a target EGR rate and an engine rotation speed.

FIG. 67 is a table characteristic diagram of a water temperature correction coefficient.

FIG. 68 is a table characteristic diagram of an atmospheric pressure correction coefficient.

FIG. 69 is a waveform chart showing an example of how each flag is set in response to a change in operating conditions.

FIG. 70 is a diagram corresponding to claims of the first invention.

[Explanation of symbols]

 4 EGR passage 6 EGR valve 41 Control unit 52 Exhaust turbine 1 Variable nozzle

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) F02D 41/14 310 F02D 41/14 310H 41/18 41/18 D 43/00 301 43/00 301N 301R 45 / 00 320 45/00 320A 340 340D F02M 25/07 570 F02M 25/07 570D 570J F term (reference) 3G062 AA01 AA05 BA04 BA06 EA04 FA15 FA16 GA01 GA05 GA06 GA08 GA13 GA15 GA21 3G084 AA01 DA04 BA07 EB20 BA21 FA00 FA07 FA10 FA11 FA12 FA13 FA18 FA20 FA33 FA37 FA38 3G092 AA02 AA10 AA17 AA18 BA01 DB03 DC06 DC09 DG06 EA01 EA18 EB01 EB03 EC02 EC05 EC07 EC08 FA06 HA01X HA01Z HA05X HA05Z HA11Z HA16X HD03 H03H03X03 HA11 HA13 JA13 LA00 LB11 LC01 LC07 MA01 MA11 NA02 NA04 NA05 NA06 NA 08 NC01 NC04 NC06 ND05 ND07 ND22 ND24 ND25 ND41 ND42 NE26 PA01Z PA16Z PB03Z PD15Z PE01Z PE04Z PE08Z PF03Z

Claims (10)

[Claims] An EGR device, means for detecting an operating condition of the engine, means for calculating a control target value of the EGR device in accordance with the detected value of the operating condition, and Means for calculating a target intake air amount or a value equivalent to a target intake air amount, means for calculating an actual intake air amount or a value equivalent to the actual intake air amount, and a value corresponding to the actual intake air amount or the actual intake air amount under learning permission conditions. Means for calculating one learning value based on an air amount error between the target value and the target intake air amount or a value corresponding to the target intake air amount; and calculating a learning value of a parameter that affects sensitivity to EGR under a learning permission condition. A learning correction coefficient in a learning value reflection permission condition based on a comparison result between a learning value of a parameter affecting sensitivity to EGR and its current value and a learning value of the air amount error. And a control unit for controlling the EGR device using the control target value of the EGR device and the learning correction coefficient under the learning value reflection permission condition. . 2. The diesel engine control device according to claim 1, wherein the parameter affecting the sensitivity to EGR is at least one of an EGR rate, a fuel injection amount, and an engine load. 3. The diesel engine control device according to claim 1, wherein the parameter affecting the sensitivity to EGR is an engine speed. 4. The EGR rate according to claim 2, wherein when the parameter affecting the sensitivity to the EGR is an EGR rate, the learned value of the EGR rate is an average value of the EGR rates under a learning permission condition. Control unit for diesel engine. 5. When the parameter affecting the sensitivity to EGR is an engine speed, the learned value of the engine speed is an average value of the engine speed under the learning permission condition. 2. The control device for a diesel engine according to claim 1. 6. The control apparatus for a diesel engine according to claim 1, wherein a condition after a lapse of a predetermined period from when the learning permission condition is satisfied is set as the learning value reflection permission condition. 7. The learning permission condition for two types of learning values, the learning value of the air amount error and the learning value of a parameter that affects the sensitivity to the EGR, is the same. Control unit for diesel engine. 8. A comparison result between a learning value of a parameter affecting sensitivity to EGR and its current value is a difference between a learning value of a parameter affecting sensitivity to EGR and its current value. The control device for a diesel engine according to claim 7. 9. When the parameter affecting the sensitivity to EGR is an EGR rate, the learning value of the EGR rate is used as the average value of the EGR rate under the learning permission condition, and the learning correction coefficient is set to a learning reflection value. When the current value of the EGR rate is smaller than the learned value of the EGR rate, the learning reflected value is set to E when the current value of the EGR rate is smaller than the learned value of the EGR rate.
9. The control apparatus for a diesel engine according to claim 8, wherein an absolute value is set to be larger than when the GR rate is larger than a learning value. 10. The diesel engine control device according to claim 1, wherein the learning correction coefficient is corrected based on a water temperature or an atmospheric pressure.